基于机器深度学习算法的城际铁路大直径盾构施工地表沉降预测  

Prediction of Ground Settlement for Large-Diameter Shield Tunnel Construction of Intercity Railway Based on Machine Deep Learning Algorithm

作　　者：孙伟良 杜守继[2] 田勇坚 崔童 Sun Weiliang;Du Shouji;Tian Yongjian;Cui Tong(Henan Railway Construction&Investment Group Co.Ltd.,Zhengzhou 450046,China;School of Naval Architecture,Ocean&Civil Engineering,Shanghai Jiao Tong University,Shanghai 200240,China;Shanghai Baoshan District Planning and Natural Resources Bureau,Shanghai 201999,China;Shanghai Xuhui District Construction and Management Committee,Shanghai 200030,China)

机构地区：[1]河南省铁路建设投资集团有限公司,河南郑州450046 [2]上海交通大学船舶海洋与建筑工程学院,上海200240 [3]上海市宝山区规划和自然资源局,上海201999 [4]上海市徐汇区建设和管理委员会,上海200030

出　　处：《石家庄铁道大学学报（自然科学版）》2022年第4期47-52,共6页Journal of Shijiazhuang Tiedao University(Natural Science Edition)

摘　　要：依托河南新郑机场至郑州南站城际铁路大直径盾构隧道施工实践,应用提出的遗传算法优化BP神经网络模型的机器深度学习算法,研究大直径盾构隧道施工中地表沉降预测规律。在盾构隧道中心线地表选取76个监测点,获取地表最大沉降数据,以前30组数据为训练数据组,对后46个测点地表最大沉降进行动态预测,并和实测值进行比较。结果表明,在46个地表最大沉降预测结果中有38个点的误差绝对值小于2 mm。应用后验差法评价动态预测结果,获得了小误差概率P值为0.87,后验差比值C值为0.59,说明提出的机器深度学习算法能够较好地实现对盾构隧道地表沉降的动态预测。Considering the issue of predicting ground settlement during the construction of the large-diameter shield tunnel,in the construction site of the intercity railway from Xinzheng Airport to Zhengzhou South Railway Station,the method of machine deep learning algorithm for optimizing the BP neural network model by genetic algorithm was proposed.The research selected 76 measuring points on the surface of the shield tunnel center line and collected the measured maximum surface subsidence data,the first 30 sets of data are used as training data groups to dynamically predict the maximum surface settlement of the last 46 measuring points.The results shows that among the 46 prediction results of maximum surface subsidence,there are 38 points whose absolute value of error is less than 2 mm.Using the posterior difference method to evaluate the dynamic prediction results,the small error probability P value is 0.87,and the posterior difference ratio C value is 0.59,which illustrates that the machine deep learning algorithm can better realize the dynamic prediction of the surface settlement of the shield tunnel.

关 键 词：机器深度学习算法 城际铁路 大直径盾构 优化BP神经网络 地表沉降预测 

分 类 号：U25[交通运输工程—道路与铁道工程]